Azure Document Intelligence REST SDK pour TypeScript

Extrayez du texte, des tableaux et des données structurées à partir de documents en utilisant des modèles prédéfinis et personnalisés.

Installation

npm install @azure-rest/ai-document-intelligence @azure/identity

Variables d'environnement

DOCUMENT_INTELLIGENCE_ENDPOINT=https://<resource>.cognitiveservices.azure.com
DOCUMENT_INTELLIGENCE_API_KEY=<api-key>
AZURE_TOKEN_CREDENTIALS=prod # Requis uniquement si DefaultAzureCredential est utilisé en production

Authentification

Important : Ceci est un client REST. DocumentIntelligence est une fonction, pas une classe.

DefaultAzureCredential

import DocumentIntelligence from "@azure-rest/ai-document-intelligence";
import { DefaultAzureCredential, ManagedIdentityCredential } from "@azure/identity";

// Dev local : DefaultAzureCredential. Production : définir AZURE_TOKEN_CREDENTIALS=prod ou AZURE_TOKEN_CREDENTIALS=<specific_credential>
const credential = new DefaultAzureCredential({requiredEnvVars: ["AZURE_TOKEN_CREDENTIALS"]});
// Ou utiliser directement une credential spécifique en production :
// Voir https://learn.microsoft.com/javascript/api/overview/azure/identity-readme?view=azure-node-latest#credential-classes
// const credential = new ManagedIdentityCredential();

const client = DocumentIntelligence(
  process.env.DOCUMENT_INTELLIGENCE_ENDPOINT!,
  credential
);

Clé API

import DocumentIntelligence from "@azure-rest/ai-document-intelligence";

const client = DocumentIntelligence(
  process.env.DOCUMENT_INTELLIGENCE_ENDPOINT!,
  { key: process.env.DOCUMENT_INTELLIGENCE_API_KEY! }
);

Analyser un document (URL)

import DocumentIntelligence, {
  isUnexpected,
  getLongRunningPoller,
  AnalyzeOperationOutput
} from "@azure-rest/ai-document-intelligence";

const initialResponse = await client
  .path("/documentModels/{modelId}:analyze", "prebuilt-layout")
  .post({
    contentType: "application/json",
    body: {
      urlSource: "https://example.com/document.pdf"
    },
    queryParameters: { locale: "en-US" }
  });

if (isUnexpected(initialResponse)) {
  throw initialResponse.body.error;
}

const poller = getLongRunningPoller(client, initialResponse);
const result = (await poller.pollUntilDone()).body as AnalyzeOperationOutput;

console.log("Pages:", result.analyzeResult?.pages?.length);
console.log("Tables:", result.analyzeResult?.tables?.length);

Analyser un document (fichier local)

import { readFile } from "node:fs/promises";

const fileBuffer = await readFile("./document.pdf");
const base64Source = fileBuffer.toString("base64");

const initialResponse = await client
  .path("/documentModels/{modelId}:analyze", "prebuilt-invoice")
  .post({
    contentType: "application/json",
    body: { base64Source }
  });

if (isUnexpected(initialResponse)) {
  throw initialResponse.body.error;
}

const poller = getLongRunningPoller(client, initialResponse);
const result = (await poller.pollUntilDone()).body as AnalyzeOperationOutput;

Modèles prédéfinis

Extraire les champs de facture

const initialResponse = await client
  .path("/documentModels/{modelId}:analyze", "prebuilt-invoice")
  .post({
    contentType: "application/json",
    body: { urlSource: invoiceUrl }
  });

if (isUnexpected(initialResponse)) {
  throw initialResponse.body.error;
}

const poller = getLongRunningPoller(client, initialResponse);
const result = (await poller.pollUntilDone()).body as AnalyzeOperationOutput;

const invoice = result.analyzeResult?.documents?.[0];
if (invoice) {
  console.log("Vendor:", invoice.fields?.VendorName?.content);
  console.log("Total:", invoice.fields?.InvoiceTotal?.content);
  console.log("Due Date:", invoice.fields?.DueDate?.content);
}

Extraire les champs de reçu

const initialResponse = await client
  .path("/documentModels/{modelId}:analyze", "prebuilt-receipt")
  .post({
    contentType: "application/json",
    body: { urlSource: receiptUrl }
  });

const poller = getLongRunningPoller(client, initialResponse);
const result = (await poller.pollUntilDone()).body as AnalyzeOperationOutput;

const receipt = result.analyzeResult?.documents?.[0];
if (receipt) {
  console.log("Merchant:", receipt.fields?.MerchantName?.content);
  console.log("Total:", receipt.fields?.Total?.content);

  for (const item of receipt.fields?.Items?.values || []) {
    console.log("Item:", item.properties?.Description?.content);
    console.log("Price:", item.properties?.TotalPrice?.content);
  }
}

Lister les modèles de document

import DocumentIntelligence, { isUnexpected, paginate } from "@azure-rest/ai-document-intelligence";

const response = await client.path("/documentModels").get();

if (isUnexpected(response)) {
  throw response.body.error;
}

for await (const model of paginate(client, response)) {
  console.log(model.modelId);
}

Créer un modèle personnalisé

const initialResponse = await client.path("/documentModels:build").post({
  body: {
    modelId: "my-custom-model",
    description: "Custom model for purchase orders",
    buildMode: "template",  // ou "neural"
    azureBlobSource: {
      containerUrl: process.env.TRAINING_CONTAINER_SAS_URL!,
      prefix: "training-data/"
    }
  }
});

if (isUnexpected(initialResponse)) {
  throw initialResponse.body.error;
}

const poller = getLongRunningPoller(client, initialResponse);
const result = await poller.pollUntilDone();
console.log("Model built:", result.body);

Créer un classifieur de document

import { DocumentClassifierBuildOperationDetailsOutput } from "@azure-rest/ai-document-intelligence";

const containerSasUrl = process.env.TRAINING_CONTAINER_SAS_URL!;

const initialResponse = await client.path("/documentClassifiers:build").post({
  body: {
    classifierId: "my-classifier",
    description: "Invoice vs Receipt classifier",
    docTypes: {
      invoices: {
        azureBlobSource: { containerUrl: containerSasUrl, prefix: "invoices/" }
      },
      receipts: {
        azureBlobSource: { containerUrl: containerSasUrl, prefix: "receipts/" }
      }
    }
  }
});

if (isUnexpected(initialResponse)) {
  throw initialResponse.body.error;
}

const poller = getLongRunningPoller(client, initialResponse);
const result = (await poller.pollUntilDone()).body as DocumentClassifierBuildOperationDetailsOutput;
console.log("Classifier:", result.result?.classifierId);

Classer un document

const initialResponse = await client
  .path("/documentClassifiers/{classifierId}:analyze", "my-classifier")
  .post({
    contentType: "application/json",
    body: { urlSource: documentUrl },
    queryParameters: { split: "auto" }
  });

if (isUnexpected(initialResponse)) {
  throw initialResponse.body.error;
}

const poller = getLongRunningPoller(client, initialResponse);
const result = await poller.pollUntilDone();
console.log("Classification:", result.body.analyzeResult?.documents);

Obtenir les informations du service

const response = await client.path("/info").get();

if (isUnexpected(response)) {
  throw response.body.error;
}

console.log("Custom model limit:", response.body.customDocumentModels.limit);
console.log("Custom model count:", response.body.customDocumentModels.count);

Motif de polling

import DocumentIntelligence, {
  isUnexpected,
  getLongRunningPoller,
  AnalyzeOperationOutput
} from "@azure-rest/ai-document-intelligence";

// 1. Démarrer l'opération
const initialResponse = await client
  .path("/documentModels/{modelId}:analyze", "prebuilt-layout")
  .post({ contentType: "application/json", body: { urlSource } });

// 2. Vérifier les erreurs
if (isUnexpected(initialResponse)) {
  throw initialResponse.body.error;
}

// 3. Créer le poller
const poller = getLongRunningPoller(client, initialResponse);

// 4. Optionnel : Surveiller la progression
poller.onProgress((state) => {
  console.log("Status:", state.status);
});

// 5. Attendre la fin
const result = (await poller.pollUntilDone()).body as AnalyzeOperationOutput;

Types clés

import DocumentIntelligence, {
  isUnexpected,
  getLongRunningPoller,
  paginate,
  parseResultIdFromResponse,
  AnalyzeOperationOutput,
  DocumentClassifierBuildOperationDetailsOutput
} from "@azure-rest/ai-document-intelligence";

Bonnes pratiques

1. Utiliser getLongRunningPoller() - L'analyse de documents est asynchrone, toujours attendre les résultats
2. Vérifier isUnexpected() - Type guard pour une bonne gestion des erreurs
3. Choisir le bon modèle - Utiliser les modèles prédéfinis quand possible, personnalisés pour les documents spécialisés
4. Gérer les scores de confiance - Les champs ont des valeurs de confiance, définir des seuils selon votre cas
5. Utiliser la pagination - Utiliser l'assistant paginate() pour lister les modèles
6. Préférer le mode neural - Pour les modèles personnalisés, neural gère mieux les variations que template